I'm a bit confused by this project. Zork's room descriptions are static, and game states are quite limited. What is the qualitative difference between this pro

I'm a bit confused by this project. Zork's room descriptions are static, and game states are quite limited. What is the qualitative difference between this project and someone manually entering the descriptions/states into an image generation AI and presenting that output as a static image file? Is it the image generation model they're presenting here? Genuine questions. I'm not trying to be dismissive. What am I missing?